Research Engineer, Post-Training

Role overview Improve voice AI models by connecting post-training research with measurable gains in production. You will investigate customer-relevant model limits, define evaluations, prepare data, fine-tune models, and deliver working improvements on a short timeline. Potential areas include supporting more languages, improving speech synthesis, separating speakers in noisy audio, and evaluating emerging voice capabilities.

Responsibilities - Find model weaknesses and set clear targets for improvement. - Develop evaluations that establish whether a model improved and help explain the result. - Curate or generate data and fine-tune models to address identified gaps. - Choose practical approaches to customer problems, including experiments with new data or external models. - Build and ship systems for production use, then assess whether they deliver measurable value. - Use customer needs and results to prioritize further research and iteration.

Requirements - Deep experience training, fine-tuning, and evaluating deep learning models. - Strong Python and PyTorch skills. - A track record of bringing machine learning systems into production. - Ability to take ownership from problem definition through measurable improvement, learning new techniques along the way. - Good judgment about which problems matter and when a solution is ready for customers. - Demonstrated work on difficult problems through projects, shipped systems, or open source contributions; a specific degree is not required.

Nice to have - Experience with speech-to-text, text-to-speech, or voice agents. - Distributed training experience. - Published research, open source work, or public benchmarks.

Benefits and work setup The role suits someone comfortable with rapid experimentation, frequent use of AI tools, and evolving priorities. Work is oriented toward delivering customer-facing improvements within weeks rather than following a highly prescriptive routine.